Errors
There are some of limitations and errors have to be pointed out. From the economic perspective, there are some more variables could effect on the rent price but have not involved in this project, such as the unit qualities (newer the property generally will rent on a higher price), as well as the unit size. I examine on the rent price difference at the community level (an area with a population from 2500 to 8000), therefore, such factors are impossible to be included. Other than that, neighborhood level criminal rate could also be a crucial factor that influences people’s decision on choosing residential location.
The result of this project also contains several errors that come from data acquisition and operation process.
Temporal errors
Most of the used data in this project are not up-to-date data and not obtained at the same time. Especially, the landuse data was from 2001. Many of the landuse type have been changed over such long time period.
Human errors
The weights of each factors derived from the pairwise comparison matrix, which are set based on my understanding to residential needs for both perspectives.
Distance errors
In Idrisi, DISTANCE module is use to calculated the Euclidean distance between features. However, in reality, travel distance is not straight line between two points. It results distance errors for every layers that have operated with DISTANCE module.
Rasterization errors
All criteria layers were converted from vector to raster format. This results in a loss of information.
Modifiable areal unit problem
This error comes from when aggregating a point-based measurement of a spatial phenomenon into census enumeration districts, such as census tract (Holt et al. 1996). It occurred in all layers that used census boundary file.
|
